International Journal of Computer Applications (0975 – 8887)
Volume 67– No.22, April 2013
Cost Effective Grading Process for Grape Raisins based
on HSI and Fuzzy Logic Algorithms
Swati P. Pawar
Amit Sarkar
SVERI’s College of Engineering, Pandharpur
Dist.Solapur State:Maharashtra
SVERI’s College of Engineering, Pandharpur
Dist.Solapur State:Maharashtra
ABSTRACT
Farmers of several progressive countries like India are
producing the grape raisins. However, the existing grading
systems in these countries are human expert based and
judgmental. The image based sorting and grading systems
developed in the advanced countries are costly and are
sometimes slow as they do analysis of individual raisin. This
work proposes to develop a cost effective grading process for
grape raisins which will give judgment about grading of bulk
of raisins sorted manually or mechanically. In this study, the
database is developed using the images taken by simple
webcam from the local raisin market. Based on the opinion of
the raisin experts, these images are grouped into four classes.
The Hue, Saturation and Intensity (HSI) color features of
these images are obtained to develop the fuzzy logic system
for the classification of the images of different grades. The
Gaussian membership function is used for developing the four
rules for four grades. The performance of the fuzzy
classification system is measured in the form of Success Rate.
The results show that for more than four features, the raisin
grading can be classified with 100 % success rates.
Keywords
Raisin sorting, Color image segmentation, Machine Vision,
Intelligent System, Feature Extraction
1. INTRODUCTION
Raisin produce in India has grown tremendously over the
period and the farmers are willing to adopt the advanced
processes for getting better yield. Existing approach of sorting
and grading of raisins is manually and with some mechanical
equipment based on size sieves. Due to manual sorting and
grading, the standardization about grades of the raisins is not
available which results in either faulty grading or exploitation
of farmers about the grading of raisins. Few studies in the
literature [1-4] have developed sorting and grading of raisins
based on individual raisin. However, these methods need
costly equipments and complicated setups. Therefore, there is
need for development of cost effective raising grading process
which is affordable for farmers as well as the small traders.
The trend of usage of image analysis and computer in
agricultural has be reviewed by Raji and Alamutu [6].They
have discussed various basic concepts and technologies
associated with computer vision, a tool used in image analysis
and automated sorting.
Computer vision for quality
inspection of fruits. Quality inspections of fruits have two
different objectives: quality evaluation and defect finding.
The machine vision captures the external feature
measurements such as color intensity, color homogeneity,
bruises, size, shape and stem identification (Lee et al. [7];
Majumdar and Jayas[8]; Shahin and Symons[9]; Paliwal et al.
[10] Shigeta et al. [11]). Several researchers have used the
advanced approaches in the process. An automatic trainable
classifier based on artificial neural network for separating
different varieties of nuts was developed Omid et al. [12].
They have concluded that the color feature the most important
parameter in classification and sorting of raisins.
Khojastehnazhand et al. [13] developed a machine vision
system to automatically determine the volume and surface
area of orange.
Recently Mollazade et al. [1] developed a process for grading
of raisins based on color and shape features for data mining of
these different classifiers used ANNs, SVMs, DTs and BNs.
Xinjie et al. [2] used least squares SVM for classification of
color and texture features for classification of raisins.
Xiaoling[3] calculated the different physical parameters of
raisin and Neural network is used for classification.
Abbasgholipour et al. [4] segmented the raisin image using
permutation coded genetic algorithm and HIS color space is
used to select desired and undesired raisin. Omid et al. [5] has
developed Microcontroller based electronic system for
grading of raisins based on their color and size. These studies
have applied these algorithms on individual raisins which is
computationally costly. However, in reality the judgmental
grading of raisins is based on the bulk quality of raisins.
In this study, a cost effective approach is developed for bulk
grading of raisins based on the color texture features extracted
using HSI color model and fuzzy logic classifier. Color
texture features are obtained through SGDM matrices where
total 21 features are obtained. Based on these features a fuzzy
classification system is developed. The Gaussian membership
function is used for the fuzzy classification system. The
performance of the system is evaluated in the form of success
rate for different feature.
Input Image
Generation of HSI color model
Texture feature calculation
Fuzzy Classifier
Class1
Class 2
Class 3
Class 4
Class 5
Fig 1: The raisin classification Process Chart
18
International Journal of Computer Applications (0975 – 8887)
Volume 67– No.22, April 2013
Class 1
Class 2
The SGDMs are represented by the function P (i, j, d, Ө)
where ‘i’ represents the gray level of the location (x, y) in the
image I(x, y), and j represents the gray level of the pixel at a
distance d from location (x, y) at an orientation angle of Ө.
Texture involves the spatial distribution of gray levels. Hence
SGDM matrices were used to calculate various texture
features. These features are useful in carrying out
differentiation algorithms for skin defect detection purpose
ahead.
For differentiating classes the following features were
calculated from the components H, S and I:
Contrast:
Ng 1
 (i  j) Pi, j 
Co 
2
i , j 1
Cluster
(1)
Shade:
Ng 1
Class 3
Class 4
Cs   i  P
x
i , j 1
Cluster
3
 j  P  P(i, j )
y
(2)
Prominence:
4
Ng 1
 (i  P  j  P ) P(i, j )
Cp 
x
i , j 1
y
(3)
Local homogeneity:
Class 5
LH 
Ng 1
1
 1  (i  j) Pi, j 
2
i , j 1
(4)
Fig 2: Typical Samples of raisin
Mean Intensity:
As shown in Figure 1 the image based raisin grading needs
system needs image acquisition equipment, image feature
extraction algorithm and image classification algorithm. In
this study, the webcam of 1.3 megapixel is used to acquire the
image database with constant height of the camera from the
raisin object. The database is developed from the local market
and the images are grouped in for classes as class1, class 2,
class 3 and class 4. The typical samples of these raisins are
shown in Figure 2. The HSI based feature extraction
algorithm is used to obtain the features of these images. The
features obtained from this algorithm are used for the
development of fuzzy classification system.
2. HSI BASED FEATURE EXTRACTION
Statistical approach, which is used here, is a quantitative
measure of arrangement of intensities in a region. Statistical
methods use second order statistics to describe the
relationships between pixels within the region by constructing
Spatial Gray-level Dependency Matrices (SGDM). Extraction
of a numerous texture features are possible using the SGDM
matrices generated in the above manner. Each pixel map is
used to generate a color co-occurrence matrix, resulting in
three CCM matrices, one for each of the H, S and I pixel
maps. These matrices measure the probability that a pixel at
one particular gray level will occur at a distinct distance and
orientation from any pixel, given that pixel has a second
particular gray level. For a position operator p, we can define
a matrix ‘Pij’ that counts the number of times a pixel with
grey-level i occurs at position p from a pixel with grey-level j.
Ng 1
 iP (i)
Mi 
i 0
x
(5)
Variance:
V
Ng 1
 (i  Mi) P (i)
x
i 0
(6)
Mean :
Mk 
1 N N
 x(i, j)
N 2 i 1 j 1
(7)
Where, P (i, j) is the (i, j) th entry in SGDM matrix and Ng is
total number of gray levels considered for generation of
SGDM matrix. Px and Py are defined by
Px 
Py 
Ng 1 Ng 1
 i  P(i, j )
i 0
j 0
Ng 1Ng 1
  jP (i, j)
i 0 j 0
19
International Journal of Computer Applications (0975 – 8887)
Volume 67– No.22, April 2013
3. FORMULATION OF FUZZY SYSTEM
A fuzzy system (FS) is any fuzzy logic (FL)-based system,
which either uses FL as the basis for there presentation of the
different forms of knowledge, or to model the interactions and
relationships among the system variables. FL and the fuzzy
sets position modeling into a novel and broader perspective in
that they provide innovative tools to cope with complex and
ill-defined systems for which classical tools are often
unsuccessful. This is because fuzzy systems are universal
approximates of nonlinear functions. Two aspects in the
design of the fuzzy system are difficult: (1) generating the
best rule set and (2) tuning the membership functions. The
rules and the membership functions must accurately capture
the relationship between the independent and dependent
variable. Unfortunately, the task of tuning of membership
function and generating rules are not independent. The task of
selecting membership functions and rule values is difficult
since the information has to be obtained from numerical data
of the system to be modeled. Another problem is selecting an
appropriate number of fuzzy sets. Most studies use experience
to come up with this number. Often, symmetric fuzzy sets are
assumed. However, assuming symmetry in the fuzzy sets also
implies assuming symmetry in the system being modeled.
In this section, we formulate the fuzzy logic system for
automatic grading of raisins the schematic representation of
fuzzy logic system is shown in fig. 3
Therefore , Rule i will be generalized as
If e1 is Ai1 AND
E2 is Ai2 AND
.
.
E7 is Ai7
THEN
Maximum success rate indicates corresponding class of raisin
Input
Output
Fuzzifier
Fuzzy Inference
Engine
Defuzzifier
Fuzzy rule base
Fig 3.Schematic representation of fuzzy logic system
4. RESULT AND DISSCUSSION
Input and output
Inputs to the fuzzy logic system are texture features obtained
from SGDM matrices the output of the system is measured in
terms of success rate the number of features are represented
by F.
F= {e1 e2 ….e7}T
D= { Each class of texture features}
Figure 1: Feature plot between mean of Hue and
Kurtosis of Saturation for 5 cluster
In this system features represent crisp numbers. Fuzzy set
with Gaussian membership function are used to define these
input variables these fuzzy sets can be defined using
following equation
 ( x)  e
 xm 2
 0.5

  
Where m is mean and
set.

(8)
is the standard deviation of fuzzy
Rules of the fuzzy system are obtained by fuzzification of
numerical values of texture features is given below
1. The fuzzy sets corresponding to each texture feature are
generated and maximum degree of membership function will
 1
be
for each fuzzy set.
2. Each texture feature is assigned to the fuzzy set with
maximum degree of membership suppose for Rule i we are
giving the inputs then fuzzy set will generate D number of
membership function A .
Figure 2:Feature plot between Standard Deviation of Hue
and Local Homogeneity of Intensity for 5 cluster
20
International Journal of Computer Applications (0975 – 8887)
Volume 67– No.22, April 2013
Figure 6:Success Rate for 5 Class
Figure 3:Feature plot between Kurtosis of Saturation and
Skewness of Intensity for 5 cluster
5. CONCLUSION
In this study a cost effective raisin grading algorithm is
developed based on the simple and low cost equipment setup
using HSI and fuzzy logic systems. The features obtained
from the HSI shows the four classes of the raisins are
separable. The features 4-7 combination shows quite good
separation. For all the features the class 1 and class2 are
slightly overlapping whereas the classes 3-4 are overlapping
to each other. The fuzzy classification system shows very
good success rates which are 100 % for all the four classes for
more than four features. This study has demonstrated the
usage of low cost camera and simple setup which will give the
grading of the raisins.
Figure 4:Feature plot between Contrast of Intensity and
Energy of Saturation for 5 cluster
Figure 5:Feature plot between Cluster Prominence oionf
Saturation and Kurtosis of saturatation for 5 cluster
21
International Journal of Computer Applications (0975 – 8887)
Volume 67– No.22, April 2013
Figure 7: Gui for feature extraction and classification
[6] A. Raji and A. Alamutu. “Prospects of Computer Vision
Automated Sorting Systems in Agricultural Process
Operations in Nigeria”. Agricultural Engineering
International: the CIGR Journal of Scientific Research
and Development”. Vol. VII. Invited Overview. February
2005.
[7] Lee, W. S., Slaughter, D. C. and Giles, D. K. . Robotic
weed control system for tomatoes. Precision Agriculture,
vol. 1, pp 95–113, 1999.
Figure 8: Grading result
6.REFERENCES
[1] K. Mollazade , M. Omid , A. Arefi “Comparing data
mining classifiers for grading raisins based on visual
features”, Journal of Computers and Electronics in
Agriculture, vol.84,pp.124-131,2012.
[2]
X. Yu, K. Liu, D.Wu and Yong He “Raisin Quality
Classification Using Least Squares Support Vector
Machine (LSSVM) Based on combined Color and
Texture Features” Journal of Food And Bioprocess
Technology,vol.5, pp.1152-1563, 2012.
[3] L.Xiaoling “Detection Level of Raisins Based on Neural
Network and Digital Image” Circuits, Communications
and System (PACCS), 2011 Third Pacific-Asia
Conference pp.1-3,july 2011.
[4] M. Abbasgholipour , M. Omid, A. Keyhani, S.S.
Mohtasebi“Color image segmentation with genetic
algorithm in a raisin sorting system based on machine
vision in variable conditions” journal of Expert Systems
with Applications,vol.38,pp.3671-3678, 2011.
[5] M. Omid, M. Sharouzi and A.R. Keyhani “Development
of an Automated Machine for Grading Raisins based on
Color and Size” Journal of Modeling and Simulation of
Systems, Vol.1,pp.157-162,2010
[8]
Majumdar, S. & Jayas, D. S. (2000). Classification of
cereal grains using machine vision II. Color models,
Transactions of the ASAE, vol.43(6), pp.1677-1680
[9] Shahin, M. A. and Symons, S. J., “A machine vision
system for grading lentils”. Canadian Biosystems
Engineering, Vol. 7,pp. 7-14, 2001.
[10] Paliwal, J., Borhan, M. S. & Jayas, D. S. Classification of
cereal grains using a flatbed scanner. Transactions of the
ASAE, pp. 036103, 2003
[11] Shigeta, K., Motonaga, Y., Kida, T. and Matsuo, M.
“Distinguishing damaged and undamaged chaff in rice
whole crop silage by image processing”, Transactions
of the ASAE Annual Meeting, pp. 043125, 2004
[12] Omid, M., Mahmoudi, A. and Omid, M. H. “ An
intelligent system for sorting pistachio nut varieties”,
Expert Systems with Applications, vol.36(9), pp.11528–
11535,2009.
[13] Khojastehnazhand, M., Omid M. and Tabatabaeefar,
A.“Determination of orange volume and surface area
using image processing technique”. International
Agrophysics, vol.23, pp.237-242, 2009.
22